The hippocampal formation plays an essential role in mnemonic processing, especially for spatial learning and memory (Amaral and Witter, 1995; Rolls et al., 2005). In human, the hippocampus seems to be important for formation of new memories involving episodic and autobiographical events, time and place, as well as for acquisition and storage of semantic memory about idea, meaning, and concept (Astur et al., 2002; Kim and Diamond, 2002; Ekstrom and Bookheimer, 2007; Hassabis et al., 2007; Chadwick et al., 2010). Thus, the hippocampus together with other medial temporal lobe structures may play a critical role in long-term declarative memory. The hippocampal neuronal circuitries include an excitatory tri-synaptic loop from dentate granule cells to CA3 and to CA1 pyramidale neurons, which further links to extrahippocampal limbic and neocortical neuronal circuitries (Frotscher et al., 2006; Eichenbaum et al., 2007; Wittner et al., 2007). This network system allows the hippocampus to participate in processing of various forms of neuronal information from multiple brain regions (Witter et al., 1989; Ishizuka et al., 1990; Lavenex and Amaral, 2000). The CA3 subregion may encord information about novel clues, sequential trials (autographical information), and arbitrary associations. It may be also critical for retrieval of cued information, pattern competition, and formation of short memory (Kesner, 2007; Bakker et al., 2008; Gilbert and Brushfield, 2009).
CA3 is often divided into CA3a-c subfields orderly from the CA2 border to the dentate hilus, with the hilar CA3 region sometimes also referred to as CA4 (Amaral and Witter, 1995). Typical CA3 pyramidal cells have a triangular or oval soma, from which the apical and basal dendrites arise and extend to the stratum lucidum and into the stratum oriens, respectively. The axons of CA3 pyramidal cells project to CA1 neurons via the Schaffer collaterals and also form recurrent innervations to other CA3 cells as well as dentate granule cells (Amaral and Witter, 1995; Scharfman, 2007; Szirmai et al., 2012). The granule cell axons, the mossy fibers, form largely axospinous synapses with CA3 pyramidal cells. Thorny excrescences are large complex spine apparatus on CA3 pyramidal cell dendrites postsynaptic to mossy fiber presynaptic terminals. CA3 pyramidal cells also receive inputs from other brain areas, including via the perforant path (Isokawa et al., 1993; Gonzales et al., 2001; Henze et al., 2002).
The morphology and anatomical connection of adult human CA3 pyramidal neurons might be comparable to that of the rodents (Seress, 2007). For instance, CA3 pyramidal neurons in the adult human hippocampus have thorny excrescence on their dendrites as seen in rodents and nonhuman primates (Lauer and Senitz, 2006). Previous studies have reported somal, dendritic, and/or axonal development of CA3 pyramidal neurons in postnatal nonhuman and human primates (Seress and Mrzljak, 1992; Seress and Ribak, 1995; Lim et al., 1997; Lauer and Senitz, 2006). Thus, in neonatal rhesus monkeys, the somata and dendrites of CA3 pyramidal cells seem comparable with that seen in the adult with respect to light microscopical and ultrastructural features. Spine development and synaptogenesis, including that of the thorny excrescences, may continue over an extended time period, that is, the first postnatal year in monkeys. To date, little is known about the early morphogenesis of human CA3 pyramidal neurons. Herein, we report a set of preliminary data on somal, dendritic, and spine development of human CA3 pyramidal neurons from midgestation to middle childhood.
MATERIALS AND METHODS
A total of 14 prenatal and 3 postnatal human brains were available at autopsy within 2 hr after death following either spontaneous or induced abortion due to maternal complications, or death caused by non-neurological diseases from three hospitals in Changsha and Xiangtang cities, Hunan province. Fetal age, defined as gestational weeks (GW), was determined based on menstrual history, body weight, and crown-rump length. Data presented in this study were from brains of five fetuses at 19, 20, 26, 35, and 38 GW, and an 8-year old (Y) child died of multiple organs failure due to cancer (Table 1). The Golgi stain in hippocampal tissues from other cases was not optimal for morphological study. Postmortem brains were used following proper consent from the patients and met all requirements and regulations set by the Ethics Committee of Central South University Xiangya School of Medicine, in compliant with the Code of Ethics of the World Medical Association (Declaration of Helsinki).
Table 1. Postmortem brains used in this study
Cause of abortion and death
Teenage pregnancy, induced abortion
Teenage pregnancy, induced abortion
Teenage pregnancy, accident fetal death
Rapid Golgi-Cox Stain
Temporal lobe blocks in ∼2 cm thickness were processed using the FD Rapid Golgi stain™ Kit (FD NeuroTechnologies, Ellicott City, MD) following the manufacturer's instruction. Tissues were rinsed briefly in double distilled water, then immersed in the Golgi impregnation solution freshly made by mixing equal volumes (at least fivefold of the tissue volume) of Solutions A and B, and stored at room temperature for 2 weeks in darkness. Tissues were then transferred into Solution C and stored at 4°C in the dark for 3 days. Solution C was changed once after the first 24 hr of immersion. After Golgi impregnation, the blocks were cut slowly into frontal sections at 100 μm thickness in a cryostat at −22°C. Every tenth of sections were collected in Solution C, mounted on gelatin-coated microslides, and air-dried in darkness. Finally, sections were dehydrated through ascending concentrations of ethanol, cleared in xylene, and coverslippered with Permount® mounting medium.
Imaging, Morphometry, and Statistical Analysis
Golgi-impregated sections were examined by an experimenter blinded to case information on a Zeiss Axioplan microscope equipped with the Neurolucida three-dimensional (3D) system (version 10) and a high-resolution motorized stage for 3D neuronal reconstruction (MicroBrightField China). Ten equally spaced (∼1 mm for the 19–26 GW cases and ∼2 mm for other cases) sections from each hemisphere were microscopically examined, and two typical CA3 pyramidal neurons in each section were automatically constructed at a magnification of 400× for quantitative analysis. The neurons selected for construction met the following criteria: (1) they were located around the middle portion of the extrahilar CA3 segment; (2) they were among those with the widest dendritic field by overall visual judgment; (3) the somata and dendritic processes were well-impregated throughout the section thickness, with no apparent truncation of the dendritic arbor; (4) they were relatively isolated from other impregated cells. Four measurements were obtained automatically by the Neurolucida software from a constructed CA3 pyramidal neuron, including somal area, and total lengths, branching nodes (or points) and spine density (per 10 μm dendritic length) of the apical and basal dendrites. A total of 20 hippocampal CA3 pyramidal neurons in each case (10 neurons per hemisphere) were subjected for the above analyses. Means and standard derivations were calculated for each set of data and compared statistically using one-way ANOVA with Bonferroni posttests (GraphPad Prism 4.02, GraphPad Software). The level for significant statistic difference was set at P < 0.05. Image illustrations were prepared with Photoshop 7.0.
Overall Morphological Development of CA3 Pyramidal Cells
In the hippocampi of the 19 GW and 20 GW fetuses, a small number of impregated cells were found in CA3 area (Fig. 1A,B). They were relatively small in size, and were round, oval, or fusiform in shape. They had no or a few but short processes. Dendritic spines were rarely identifiable on the labeled cells. We did not perform dendritic morphometry for these two cases because (1) their dendritic structures (branches and spines) were not evident microscopically; and (2) the Neurolucida analysis failed to obtain sufficient dendritic measurements for a systematic comparison with other cases.
In the 26 GW hippocampus, impregated CA3 cells seemed slightly larger in somal size relative to 19 and 20 GM (Fig. 1A–C). The somata of the labeled cells were mostly triangular or pyramidal. A relatively long apical dendrite and a few short basal dendrites were seen on many of the labeled somata, both having a few orders of branches (Fig. 1C, arrowhead). Spine structures were found on the somata as well as the apical and basal dendrites at high magnifications (Fig. 2A–C).
In the hippocampi from the 35 GW, 38 GW, and 8 Y cases, a large number of CA3 pyramidal cells were stained, with apical and basal dendrites clearly seen but extended over long distances in the strata radiatum, lucidum, and oriens (Fig. 1D–F). The apical dendrites were considerably thick and had several orders of branches over the stratum radiatum (s.r.) and s.l. (Fig. 1E,G). Basal dendrites arose from the low portion of the somata, and were relatively thin, ranged from two to several and distributed in the strata pyramidale and oriens, some of which gave rise to multiple branches (Fig. 1G,H). An axon or axon-like process could be identified on some pyramidal cells, often arising from the midpoint of the somal base (Fig. 1G). Spines occurred commonly on the apical and basal dendrites, as well as the somata (Fig. 2G). In the hippocampus of the 8 Y case, heavily branched apical and basal dendrites were found on CA3 pyramidal cells, which appeared to be densely packed with spines (Figs. 1F,H and 2F,G).
The dendritic spines on the impregated CA3 pyramidal cells exhibited variable morphology. Most spines were close to the dendritic shaft or somal surface and had a short neck, while a small of them appeared to be long-neck spines (Fig. 2). In the 26 GW case, a few spines appeared to have an enlarged and somewhat branched spine head (Fig. 2C). These relatively large spines were more frequently seen on CA3 pyramidal neurons in the 35 GW and 38 GW cases (Fig. 3A–C). In the 8 Y case, spines were densely packed on the apical and basal dendrites (Figs. 2F,G and 3C). Large complex spines characteristic of thorny excrescence were seen in the three old cases (Fig. 3D). These complex spines were found largely on the proximal portion, that is, the trunk and the first 1–2 orders, of the branches of the apical dendrites (Fig. 3A–C). Especially in the 8 Y case, thorny excrescences were fairly abundant, covering a large part of the apical dendrite and also occurring on some basal dendrites (Fig. 3D).
Quantitative Developmental Data of CA3 Pyramidal Cells
3D reconstructions of the CA3 pyramidal cells and their processes were created by the Neurolucida software (Fig. 4A,B). The cell drawings displayed a trend of age-dependent morphogenesis that was most prominent in the dendritic components, especially after 26 GW. Thus, the dendrite tree of CA3 pyramidal neuron was relatively simple in the 26 GW hippocampus, with 0–4 basal dendrites illustrated (mean = 2). The dendritic arborization became fairly complex at 35 GW, with the number of basal dendrites reached 2–5 (mean = 3.5). The number of the primary basal dendrites ranged from 4 to 8 in the 38 GW (mean = 4.5) and 8 Y (mean = 5.4) cases (Fig. 4A). Quantitative data derived from the programmed measurements of the CA3 pyramidal cells included somal size, total dendritic length (not shown), and the total lengths of the apical and basal dendrites, as well as spine density in the four cases from 26 GW to 8 Y. Means of these measurements were calculated and plotted as a function of age, and subjected to statistical analyses (Fig. 4B–E).
The average somal area of the 20 quantified CA3 pyramidal cells was 236.8 ± 81.4 μm2 for the 19 GW fetus, 271.7 ± 88.6 μm2 for the 20 GW fetus, 363 ± 105.9 μm2 for the 26 GW fetus, 474.3 ± 115.5 μm2 for the 35 GW fetus, and 449.5 ± 116.8 μm2 for the 38 GW fetus (Fig. 4B). The largest mean somal size was found in the 8 Y case, reached to 656.8 ± 125.7 μm2. One-way ANOVA test indicated that there was age-dependent statistically difference in somal size (P < 0.0001, F = 41.32, df = 3.76). Bonferroni post hoc analysis showed statistically significant difference for the 19 and 20 GW cases relative to the 35 GW, 38 GW, and 8 Y cases (P < 0.01 to P < 0.001), for the 26 GW case relative to the 35 GW (P < 0.05) and 8 Y (P < 0.001) cases, and also for the 36 GW and 38 GW relative to the 8 Y hippocampi (P < 0.001).
The mean total length of apical dendrites was 146.0 ± 65.6 μm for the 26 GW fetus, 1445.6 ± 586.8 μm for the 35 GW fetus, 2329.0 ± 815.6 μm for the 38 GW fetus, and 2397.0 ± 250.7 μm for the 8 Y subject (Fig. 4C). Thus, the apical dendritic length increased significant from 26 GW to 8 Y (P < 0.0001; F = 81.80, df = 3.76). Bonferroni's Multiple Comparison Test indicated statistically significant difference between each pair of the four age points (P < 0.001), except for the 8 Y relative to the 38 GW cases (P > 0.05). The mean total length of the basal dendrites was 248.5 ± 103.5 μm for the 26 GW fetus, 1720.0 ± 795.2 μm for the 35 GW fetus, 1818.6 ± 565.5 μm for the 38 GW fetus; and 2076.1 ± 664.9 μm for the 8 Y subject (Fig. 4C). There was significant difference in the total length of basal dendrites among the four cases (P < 0.0001; F = 38.78, df = 3.76). Bonferroni's Multiple Comparison Test indicated statistically significant difference between the 26 GW case relative to each of the remaining cases (P < 0.001). No difference existed between the 35 GW, 38 GW, and 8 Y cases (P > 0.05).
The mean of the apical dendrite branching points or nodes was 4.0 ± 1.1 for the 26 GW fetus, 8.9 ± 4.2 for the 35 GW fetus, 17.8 ± 8.5 for the 38 GW fetus; and 12.4 ± 6.1 for the 8 Y subject (Fig. 4D). There was significant difference among the four cases (P < 0.0001; F = 21.10, df = 3.76). Bonferroni's Multiple Comparison Test indicated statistical significant difference between each pair of cases (P < 0.05 to P < 0.01), except for the 35 GW fetus relative to the 8 Y subject (P < 0.001). The mean of the basal dendrite nodes was 3.3 ± 2.1, 22.4 ± 6.3, 23.3 ± 8.2, and 20.0 ± 3.6, for the 26 GW, 35 GW, 38 GW, and 8 Y hippocampus, respectively (Fig. 4D). There existed an age-related difference among the cases (P < 0.0001; F = 57.02, df = 3.76), with post hoc test indicated statistically significant difference between the 26 GW case relative to each of the remaining cases (P < 0.001), whereas no difference between the 35 GW, 38 GW, and 8 Y cases (P > 0.05).
Dendritic spine densities.
The mean spine density (expressed as number of spines per 10 μm dendritic length) on the apical dendrites was 2.1 ± 0.8, 3.8 ± 0.6, 5.7 ± 0.8, and 7.0 ± 1.6, for the 26 GW, 35 GW, 38 GW, and 8 Y hippocampus, respectively (Fig. 4E). Thus, there existed an age-related increase of spine density on the apical dendrites of the CA3 pyramidal neurons (P < 0.0001; F = 84.99, df = 3.76). Bonferroni's Multiple Comparison Test indicated statistically significant difference between any two comparing cases (P < 0.001). The mean spine density on the basal dendrites was 1.6 ± 0.8, 3.5 ± 0.5, 5.1 ± 1.0, and 7.3 ± 0.5, for the 26 GW, 35 GW, 38 GW, and 8 Y hippocampus, respectively (Fig. 4E). As with apical dendritic spines, the density of the basal dendritic spines increased age-dependently (P < 0.0001; F = 215.5, df = 3.76). Also, there was statistically significant difference for a given case relative to another among the prenatal and postnatal hippocampi (P < 0.001 for all paired mean comparison).
Understanding neuronal morphogenesis in the human hippocampal formation, including in CA3, is of considerable importance given its relevance to cognitive function and its modulation during development and at adulthood (Astur et al., 2002; Sandi et al., 2003; Kesner, 2007; Rolls et al., 2005; Gilbert and Brushfield, 2009). Dendritic/spine alteration and aberrant synaptoplasticity in CA3 pyramidal neurons are reported in various developmental, neurological, and age-related cognitive disorders (Drakew et al., 1996; Apostolova et al., 2006; Conrad, 2006; Kolomeets et al., 2007; Tsamis et al., 2010). Specifically, pre- and postnatal factors can affect the morphogenesis of CA3 pyramidal neurons, potentially leading to long-term cognitive deficits (Sunanda et al., 1995; Bartesaghi and Severi, 2002; Hosseini-Sharifabad and Hadinedoushan, 2007; Jia et al., 2010; Bock et al., 2011; Teicher et al., 2012). Using Golgi impregnation and Neurolucida technique, we show here a preliminary, but the first, set of data on somal and dendritic development of human CA3 pyramidal neurons. As only a small number of cases were available, it is expected that our data may reflect a certain trend of developmental process, but there exists substantial individual variability.
Somal and Dendritic Growth of Human CA3 Pyramidal Neurons
In rhesus monkey with a gestation period of 165 days, the CA3 pyramidal neurons are generated during embryonic day 65 (E65) to E70 (Rakic and Nowakoski, 1981). By extrapolation, human (pregnancy time 266 days) CA3 pyramidale neurons would be probably generated around E105–E113, or during 13 GW to 14.1 GW. Indeed, cell proliferation in the hippocampal ventricular zone is rare after 24 GW in humans (Seress et al., 2001). In this study, Golgi stained pyramidal-like neurons are readily present in CA3 in the 19 GW and 20 GW (midgestation) fetuses, although they often have an oval soma with rare neuronal processes. By 26 GW, the somata are often triangular, with apical and basal dendrites distinguishable on the somata. The mean somal size of CA3 pyramidal neurons nearly doubles from midgestation to term (38 GW) and might further increase to a certain extent at least until 8 Y.
The Neurolucida quantification suggests that dendritic arborization and branching of the human CA3 pyramidal neurons progress largely after 26 GW. In the 26 GW fetal hippocampus, only a few but short dendritic processes are identifiable on CA3 pyramidal cells. In near-term fetuses (35–38 GW), the mean total length and number of branching notes of the basal dendrites are greatly increased relative to 26 GW (e.g., to almost eightfold for dendritic length). However, these basal dendrite measurements are comparable between the 38 GW and 8 Y cases. The mean total length and number of branching nodes of the apical dendrites also increases dramatically from 26 GW to 38 GW, with the former seemingly not further increasing after birth. It should be noted that the data suggest a slight but statistically significant reduction in the number of branching nodes on the apical dendrites in the 8 Y child relative to the 38 GW fetus. It is unclear if this finding reflects a certain postnatal trend of dendritic reorganization; or it is due to the use of a single child case available for morphometry in this study.
In a previous study by Seress and Ribak (1995), the somata and dendrites of CA3 pyramidal cells in newborn rhesus monkey are found to be similar to those of adults at light and electron microscopic levels. The findings in this study also suggest a relatively “advanced” maturation of human CA3 pyramidal neurons around term with regard to somal growth, dendritic elongation, and branching. Together, it appears that the basic somal and dendritic morphology is largely mature-looking by birth in both non-human and human primates (Seress and Ribak, 1995).
Spine Development on Human CA3 Pyramidal Neurons
In rhesus monkey, the overall spine density of the CA3 pyramidal neurons increases during the first year of life (Seress and Ribak, 1995). Thorny excrescences are detectable at birth on the proximal segments of apical and basal dendrites of CA3 pyramidal neurons and hilar mossy cells in nonhuman and human primates (Seress and Mrzljak, 1992; Seress and Ribak, 1995; Lim et al., 1997; Lauer and Senitz, 2006). In human, the first large, complex excrescences appear on the proximal dendrites of hilar mossy cells around 7 month of age, whereas adult-like excrescences are seen on these cells by 5 year of age (Seress and Mrzljak, 1992).
Herein, we show that dendritic spines are clearly identifiable on Golgi-impregnated CA3 pyramidale neurons in human hippocampus at 26 GW. The mean spine densities on the apical and basal dendrites are increased to ∼threefold around term and may continue to rise after birth. Thus, spine density on the basal dendrites is increased to 1.5 fold at 8 Y relative to 38 GW, with a lesser but also significant elevation for that on the apical dendrites. These data suggest that spine genesis on human CA3 pyramidal neurons occurs largely during the last trimester, but a trend of increase continues postnatally at least until 8 year of age. This Golgi study also shows the occurrence of thorny excrescences on human CA3 pyramidal neurons at relatively late gestational stages, that is, in near-term fetuses. Thus, a few small-sized cluster spines are found on the apical dendrites in the 35 GW and 38 GW cases. In the 8 Y case, a large number of thorny excrescences are clearly present on the proximal apical and basal dendrites, which appear to be morphologically comparable with that reported in the adult human CA3 (Lauer and Senitz, 2006). Overall, data from this study and the earlier monkey and postnatal human hippocampal studies point to an extended course of dendritic spine development and perhaps synaptogenesis/plasticity on nonhuman and human CA3 pyramidal neurons (Seress and Mrzljak, 1992; Seress and Ribak, 1995).
In summary, this quantitative Golgi study shows that somal growth and dendritic arborization of human CA3 pyramidal neurons occur largely after midgestation, with a mature-looking morphology achieved around term. Spine development proceeds during the third trimester and appears to continue during childhood. Thorny excrescences on CA3 pyramidal neurons are likely developed mostly after birth, and by 8 year of age, they appear to be morphologically comparable with those described in the adult human hippocampus. Together, our data are consistent with a notion that late gestational stage and early childhood may be a vulnerable period in human during which exposure to certain risk environmental factors, such as alcohol, malnutrition, and stress (Kim and Diamond, 2002; Conrad, 2006; Frotscher et al., 2006; Teicher et al., 2012), may lead to impaired spine formation and synaptic circuitry development in human hippocampal formation.
The authors thank Professor Kwok-Fai So at the Department of Anatomy of The University of Hong Kong for providing the Neurolucida facility for image analysis in this study.